Norlida, Jamil (2023) Indirect monitoring of surface quality based on the integration of support vector machine and 3D I-kaz techniques in the machining process. PhD thesis, Universiti Malaysia Pahang Al-Sultan Abdullah (Contributors, Thesis advisor: Ahmad Razlan, Yusoff).
|
Pdf
ir.NORLIDA BINTI JAMIL_PMF16001.pdf - Accepted Version Download (393kB) | Preview |
Abstract
Improved machining process quality can contribute to sustainable manufacturing in terms of economic, environmental, and social sustainability. Reducing waste, increasing efficiency, and improving product quality can also help manufacturers to reduce costs and increase productivity rate. Machining is one of the common methods in industry and plays a central role in modern manufacturing. For many years, researchers have been studying monitoring methods to produce the best surface quality. The measurement involves three distinct techniques, which are categorised into quantitative and visualisation methods. Monitoring methods can be classified as either direct or indirect methods. The common method of measuring machining quality undergoes manufacturing bottlenecks, as it is constrained by human inspection and expensive equipment. A slow process leads to higher labour costs and a high risk of equipment damage to the workpiece. The present study aims to bridge this gap by leveraging the capabilities of 3D I-kaz and medium Gaussian SVM models to improve accuracy and classification rates for determining surface quality. The specific objectives are to analyse the impact of machining parameters on statistical analysis, classify acceleration signals for surface roughness identification using SVM, integrate SVM with 3D I-kaz to improve surface quality identification and validate its effectiveness through experiments. The quantification of signal processing for ductile iron, FCD450 material on cutting parameters: rotation speed with 1000–3026 rev/mm, feed rate of 120–720 mm/min, axial of 0.75–3.5 mm, and radial depth of cut (RDOC) is studied and validated through experiments under dry and minimum quantity lubrication (MQL) conditions. Surface roughness was measured to verify the acceleration signal, while Pearson’s correlation coefficient was used to evaluate the correlation strength between the acceleration signal and surface roughness. The calculated coefficient, r-value, was found to be 0.6543, which indicates a positive but nonlinear correlation between the acceleration signal and surface roughness. The kurtosis value measured from acceleration signals and surface roughness information was then used to classify the machining condition and identification of the surface quality. In the first experiment, the model displayed an accuracy of 84.87% and 84.57% in terms of F1 values. It was observed that by adjusting the hyperparameter, the model’s accuracy was augmented to 85.53% and its F1 score was enhanced to 84.93%. Additionally, the model was applied in the second experiment, resulting in an accuracy of 84.0%. Before the classification of machined surface condition, the condition is identified through the support vector machine (SVM) technique, and it was demonstrated that the condition could be demarcated into five different levels of surface quality. From the experimental test data, acceleration and average roughness (Ra)-based indicators are identified for correlation analysis. A relation is developed, which enables the prediction or identification of surface quality directly based on the selected based indicators (3D I-kaz coefficient) without having to inspect the milling process for surface roughness. It was demonstrated that the integration of the 3D I-kaz and SVM model resulted in an accuracy and F1 score of 96.0% and 96.3% respectively, suggesting that the quantification data is viable for surface quality identification. A monitoring experiment was conducted in this study to validate the identification of surface quality through the instantaneous surface roughness level obtained from the experiment. In conclusion, indirect monitoring of surface quality using vibration signals can quickly identify the surface quality using SVM and 3D I-kaz analyses, thus reducing the time and cost associated with manual inspection and allowing for its use in many other machining processes.
| Item Type: | Thesis (PhD) |
|---|---|
| Additional Information: | Thesis (Doctor of Philosophy) -- Universiti Malaysia Pahang – 2023, NO. CD: 13476, SV: Prof. Ir. Dr. Ahmad Razlan Bin Yusoff |
| Uncontrolled Keywords: | support vector machine (SVM) technique |
| Subjects: | T Technology > TA Engineering (General). Civil engineering (General) T Technology > TS Manufactures |
| Faculty/Division: | Institute of Postgraduate Studies Faculty of Manufacturing and Mechatronic Engineering Technology |
| Depositing User: | Mr. Nik Ahmad Nasyrun Nik Abd Malik |
| Date Deposited: | 06 Jun 2024 04:10 |
| Last Modified: | 06 Jun 2024 04:10 |
| URI: | http://umpir.ump.edu.my/id/eprint/41485 |
| Download Statistic: | View Download Statistics |
Actions (login required)
 |
View Item |